← Back
Optimized Retrieval-Augmented Generation (RAG) System for Enterprise Knowledge Bases

Optimized Retrieval-Augmented Generation (RAG) System for Enterprise Knowledge Bases

Python · FAISS · LlamaIndex · HuggingFace · Sentence-Transformers · Cross-Encoder Reranking · Streamlit

Highlights

  • Built an end-to-end RAG pipeline over 6,000+ documents, enabling grounded enterprise QA with citation support.
  • Conducted embedding and retrieval ablations (MiniLM, BGE, E5, chunking, depth), improving answer accuracy by 10%.
  • Reduced hallucination by 15% via hybrid retrieval and cross-encoder reranking while maintaining 1.6s p95 latency; implemented evaluation framework measuring Recall@k, MRR, and grounded accuracy.

Tech stack

PythonFAISSLlamaIndexHuggingFaceSentence-TransformersCross-Encoder RerankingStreamlit

Next steps

Want this page to look even stronger? Add: architecture diagram, metrics (accuracy/latency), screenshots, and links. Update data/profile.ts to include your GitHub/demo URLs.